How to run Whisper locally in Python

Setup

pip install openai-whisper

import whisper

Examples

Integration steps

1. Install the openai-whisper package via pip.
2. Import the whisper module in your Python script.
3. Load a Whisper model locally using whisper.load_model("base").
4. Call model.transcribe() with the path to your audio file.
5. Extract the 'text' field from the transcription result.
6. Use or save the transcribed text as needed.

Full code

import whisper

# Load the Whisper model locally
model = whisper.load_model("base")

# Transcribe an audio file
result = model.transcribe("audio.mp3")

# Print the transcribed text
print("Transcription:", result["text"])

Transcription: This is a sample transcription of the audio file.

API trace

N/A: local model inference with audio file path input

{"text": "transcribed text", "segments": [...], "language": "en"}

Variants

from faster_whisper import WhisperModel

model = WhisperModel("base", device="cpu")
segments, _ = model.transcribe("audio.mp3")
text = "".join([segment.text for segment in segments])
print("Transcription:", text)

import whisper
import asyncio

async def transcribe_async():
    model = whisper.load_model("base")
    result = await asyncio.to_thread(model.transcribe, "audio.mp3")
    print("Transcription:", result["text"])

asyncio.run(transcribe_async())

Performance

• Use smaller Whisper models (tiny, base) for faster transcription.
• Preprocess audio to reduce length or sample rate to speed up.
• Batch multiple files if using async to maximize throughput.